David HouandGitHub 343b256deb PoC fast winograd compile (#1771) ...

* proof of concept for variable replace global load

* small hacks to make faster

* clean up a little?

* linter

* allow substituting with an expression

* clean up a little

* fix everything

* try to fix bug?

* type annotation

* typing

* typing

2023-09-05 21:14:40 -07:00

..

codegen

PoC fast winograd compile (#1771)

2023-09-05 21:14:40 -07:00

nn

perf: use enumerate where possible (#1692)

2023-08-30 10:41:51 -07:00

renderer

Parens and gls (#1768)

2023-09-04 16:09:01 -07:00

runtime

add 2 to locals, uops debug 5 (#1782)

2023-09-05 19:44:43 -07:00

shape

PoC fast winograd compile (#1771)

2023-09-05 21:14:40 -07:00

graph.py

[ready] avoid in-function graph imports in lazy.py (#1666)

2023-08-25 13:56:28 -04:00

helpers.py

perf: faster and more readable merge_dicts (#1775)

2023-09-05 14:42:19 -07:00

jit.py

llama JIT python runtime speedup (#1633)

2023-08-30 07:51:05 -07:00

lazy.py

perf: create buffer only when needed (#1684)

2023-09-02 17:43:29 -07:00

mlops.py

UnaryOps.NEG (#1749)

2023-09-03 12:44:26 -07:00

ops.py

Parens and gls (#1768)

2023-09-04 16:09:01 -07:00

tensor.py

fix attn_mask None issue (#1786)

2023-09-05 20:45:54 -07:00